Automated karyotyping is an important procedure in cytogenetics labs worldwide. Multiplex fluorescence in situ hybridization (M-FISH) is a relatively recent development that uses multicolor chromosome painting probes and multispectral image analysis to identify subtle and complex chromosomal rearrangements. It promises to make automated karyotyping faster, more accurate, and easier to interpret, both in clinical situations and in cancer research. The major factor limiting the ability of M-FISH instruments to resolve the chromosomal origin of the DNA in abnormal chromosomes is pixel classification accuracy. The goal of this project is to develop improved software techniques to improve significantly the accuracy of pixel classification in M-FISH systems, thereby maximizing the impact of this important new technology on the practice of cytogenetics. The instruments that are commercially available now implement only relatively rudimentary pixel classification algorithms for identifying the homologue origin of chromosomal DNA. In this project, we apply state-of-the-art pattern recognition techniques to M-FISH to improve pixel classification accuracy far beyond what is currently offered by commercial systems. The improved M-FISH system that will result from this research will automatically find and flag both subtle and complex structural abnormalities (insertions and translocations of genetic material) with high accuracy, thereby assisting both cancer research and genetic diagnosis. Having established feasibility of several algorithmic innovations in Phase I, we plan, in Phase II, to develop these improvements fully and test the system in routine clinical use on prenatal, postnatal and cancer specimens from amniocentesis, peripheral blood, and bone marrow. The enhancements will be integrated into commercially available M-FISH instruments. The resulting second generation commercial instruments will be superior to currently available systems for elucidating structural rearrangements within chromosomes. Certain of the techniques will be published as well, to the benefit of others using M-FISH.